Building explainable machine learning models by design

Building responsible #machinelearning applications is very important especially when these applications are used in high-risk sectors such as healthcare, judiciary and finance.

State-of-the-art (SOTA) machine learning algorithms such as Random forest and Boosted trees are best in class when it comes to model performance but fall behind linear models when it comes to model explainability. The approach used by most machine learning practitioners to fill-in the explainability gap of SOTA models consist of using tools and techniques such as LIME, SHAP and sensitivity analysis to explain these models.

Explainable Boosting Machine ( EBM) is a Generalized Additive Model ( GAM ) with accuracy comparable to SOTA models and enjoy the interpretability which is associated to GAMs.

Check out the links below for more information.

Interpret ML: InterpretML

Explainable Boosting Machine: Explainable Boosting Machine — InterpretML documentation